In a transfer geared toward propelling India and serving to native startups in AI innovation, Union Minister Rajeev Chandrasekhar unveiled a plan to determine a cluster of 25,000 GPUs. 

This initiative, set to be realised via a public-private partnership (PPP), is in dialogue on the highest ranges of the Ministry of Electronics and IT. Chandrasekhar introduced this bold endeavour in September, shedding mild on a dedication to fostering real-world AI functions. 

The minister said that the continuing dialogue about AI is nearly at all times about functions like ChatGPT. “Our mission is real-world AI use cases. We are looking at health, governance, education and creating AI-specific integrated circuits for those applications,” he mentioned in an announcement.

Abundant Data, Need GPUs

At its core, this initiative isn’t just about enhancing India’s AI prowess; it’s additionally about safeguarding the nation’s information sovereignty. The shortage of GPUs throughout the nation has pushed many companies to depend on abroad cloud-based options. Recognising the urgency of addressing this concern, one of the seven AI working groups established by MeitY strongly advisable the creation of a 25,000 GPU cluster.

Experts contend that India’s considerable information assets and human capital necessitate supercomputing energy to compete successfully within the international AI enviornment. This large cluster of GPUs is a pivotal step towards attaining this goal. For context, India’s present quickest supercomputer, ‘Airawat,’ boasts a mere 640 GPUs, ranking 75th globally. In distinction, the world’s prime supercomputers function over 30,000 GPUs.

Once the proposal is finalised, the federal government will provoke a normal tendering course of to ask non-public firms to take part in establishing the GPU cluster. Notably, discussions with tech giants like NVIDIA, together with a gathering between its founder and CEO Jensen Huang and Prime Minister Modi, have underscored the potential for collaboration. 

Huang advised AIM that India will get about 10s of thousands of GPUs to be able to construct infrastructure – i.e. about 1,00,000 GPUs. “We are going to bring out the fastest computers in the world. These computers are not even in production [so far]. India will be one of the first countries in the world [to get them],” Huang mentioned, confirming that these can be quicker than something the world has ever seen.

Going by the figures that the NVIDIA founder was hinting at, the 25k GPU cluster might very nicely be half of a bigger cargo containing 1,00,000 GPUs.

However, whereas this preliminary initiative is undeniably important, it merely marks India’s preliminary foray into the sector of nations selling AI analysis and growth capabilities. Compared to firms like OpenAI, which possess over 20,000 GPUs, and a $10 billion funding from Microsoft, the federal government would require private-sector partnerships to completely harness the potential of this computing energy.

The estimated price of this bold venture falls within the vary of INR 8,000-10,000 crore and is at the moment beneath deliberation on the highest echelons of Meity. Indian AI startups, business gamers, and outstanding CEOs have persistently advocated for such investments in computing capability to deal with the shortage and prohibitive price of GPUs.

Issues Facing Local AI Advancement

National initiatives to assemble supercomputers and tasks geared toward coaching LLMs in a number of Indian languages are already in progress. However, there are numerous points dealing with it.

When firms search entry to GPUs from cloud service suppliers or GPU producers, they usually face extended wait times, generally spanning months. To deal with this bottleneck, firms are urging the federal government to put money into important computing infrastructure for AI techniques and functions. Without this help, India dangers lagging within the international AI race, which encompasses functions from banking to house stations, all powered by algorithmic intelligence.

“Leading Indian startups are grappling with the problem of acquiring entry to 1,000-GPU clusters, usually diverting worthwhile funds from their fundraising efforts, MD of PeakXV Partners(beforehand Sequoia India) Rajan Anandan, said; emphasising the necessity for inexpensive entry to such clusters, suggesting a pyramid method: free entry for educational establishments, sponsored entry for startups, and business entry for bigger firms.

IBM CEO Arvind Krishna, not too long ago additionally reiterated the identical saying, “In many nascent technologies, you often need the government to step in first before others will follow.”

“The government should set up a national Al computing centre,” Krishna confused.

India is house to over 60 energetic genAI startups as of May 2023, having obtained roughly $475 million in funding between 2021 and 2023. While India’s AI ecosystem is flourishing, it lags behind nations just like the United States and Israel in foundational AI fashions and funding. 

What About Other Countries?

In distinction, governments in different nations have dedicated substantial funds to safe GPU entry for analysis functions. The UK, Saudi Arabia, the UAE, and Chinese tech giants have all invested considerably in buying GPUs to bolster their AI capabilities. Even the United States has supplied a 50% low cost to researchers engaged on supercomputing tasks.

Lack of native GPU entry could drive Indian firms to go for international cloud suppliers, resulting in information localisation points. Sudipta Ghosh, Partner and Leader of Data and Analytics at PwC India, emphasised the significance of regulatory frameworks for accountable and moral AI. Such frameworks wouldn’t solely bolster public belief but additionally guarantee transparency and accountability.

The shortage of GPUs has additionally made them dearer in India, main suppliers to be cautious about transport them to the nation the place demand, fee capabilities, and ticket sizes are comparatively smaller.

Addressing this problem could require home GPU manufacturing, probably supported by authorities incentives. Prashant Garg, Partner at EY, suggests following the mannequin of attracting international automakers to arrange store in India.

Efforts are already underway to offer managed GPU access to AI startups via collaborations between Nasscom and the Centre for Development of Advanced Computing (CDAC). However, consultants agree that India must put money into elementary analysis and entice prime AI scientists to drive innovation.

Simultaneously, collaborations between American GPU producer NVIDIA and Indian giants Reliance Jio and Tata Sons maintain promise for offering computing infrastructure to rising companies.

Looking forward, loads of curiosity from different semiconductor giants like AMD, Micron, SOLIS-IDC, Foxconn, STMicroelocronics might snowball into loads of GPU producers coming in as nicely. With easy laws and beneficial enterprise situations, India might have GPUs being manufactured domestically.